Electronic Device, Action Instance Generation Method, and Recording Medium

4. The electronic device of claim 1, wherein by executing the at least one instruction, the processor sets a relatedness score, which is of a location of one of a snippet, a preceding snippet, and a following snippet in which a probability of including an action instance is greater than or equal to a first threshold probability, to be relatively high on the basis of the feature information corresponding to the video.

5. The electronic device of claim 1, wherein by executing the at least one instruction, the processor is configured to set a starting score, which is of a location of one of a snippet and a preceding snippet in which a probability of being a starting point of an action instance is greater than or equal to a second threshold probability on the basis of the feature information corresponding to the video.

6. The electronic device of claim 1, wherein by executing the at least one instruction, the processor is configured to set an ending score, which is of a location of one of a snippet and a following snippet in which a probability of being an ending point of an action instance is greater than or equal to a third threshold probability on the basis of the feature information corresponding to the video.

8. The electronic device of claim 7, wherein the at least one time section may include at least one of a first time section corresponding to a combination of snippets in which a sum of the starting score and the ending score is relatively high, a second time section corresponding to a combination of snippets in which the relatedness score is greater than or equal to a threshold value, and a third time section corresponding to a weighted sum of the first time section and the second time section.

14. The method of claim 12, wherein the calculating comprises setting a relatedness score, which is of a location of one of a snippet, a preceding snippet, and a following snippet in which a probability of including an action instance is greater than or equal to a first threshold probability, to be relatively high on the basis of the feature information corresponding to the video.

15. The method of claim 12, wherein the calculating comprises setting a starting score, which is of a location of one of a snippet and a preceding snippet in which a probability of being a starting point of an action instance is greater than or equal to a second threshold probability, to be relatively high on the basis of the feature information corresponding to the video.

16. The method of claim 12, wherein the calculating comprises setting an ending score, which is of a location of one of a snippet and a following snippet in which a probability of being an ending point of an action instance is greater than or equal to a third threshold probability, to be relatively high on the basis of the feature information corresponding to the video.

18. The method of claim 17, wherein the at least one time section includes at least one of a first time section corresponding to a combination of snippets in which a sum of the starting score and the ending score is relatively high, a second time section corresponding to a combination of snippets in which the relatedness score is greater than or equal to a threshold value, and a third time section corresponding to a weighted sum of the first time section and the second time section.